Visual positioning method for complex road environment in automatic driving scene

• Main Authors: XIA QIANGQIANG, WU XUDONG, SUN HONGJIAN, WU YONG, WANG KAI, ZHAI XIANXU, GU TAOTAO, LIU HENG
• Format: Patent
• Language: Chinese; English
• Published: 15.03.2024
• Subjects: CALCULATING; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The invention belongs to the technical field of automatic driving, and discloses a visual positioning method for a complex road environment in an automatic driving scene, which comprises the following steps of: 1, eliminating dynamic feature points of an image transmitted by a camera sensor through a Lite DETR network and an LK optical flow method based on a Gaussian mixture model clustering method, and then carrying out feature matching and pose estimation; and 2, trajectory optimization is carried out at the rear end by adopting a beam adjustment optimization method. And step 3, extracting the feature vector of the current frame through a ViT-Base network, carrying out binarization processing on the feature vector of the current frame and the feature vector of the key frame, and carrying out robust loopback detection based on a Hamming distance. And step 4, establishing a motion trail diagram according to the image depth information. According to the invention, complex road environment conditions can be bet